Garbage In, Garbage Out: AI Content Lessons for SMM

I once had a client who fed their product description into ChatGPT and asked for Instagram captions. The AI spat out a glowing recommendation for a competitor's product. The client hit publish without reading it—and spent the next week doing damage control.

That's exactly what Margaret Atwood experienced when she asked Anthropic's Claude about the TV show Father Brown. The chatbot gave her a confident wrong answer. She called it lying, but really it's just the math doing its thing. For social media managers, this story is a flashing red warning light.

Why Atwood's Claude Experiment Matters to Your Content Strategy

You might be thinking: "But AI tools save me hours every week!" Here's the short answer: yes, they do—but only if you treat them like a junior intern who always sounds convincing but sometimes invents facts. The moment you stop double-checking, you're one click away from a PR crisis.

We tested this with our own social posts. We asked an AI to write a tweet about a new feature, and it added a fake statistic. The tweet went through our review process, but if it hadn't, we would have shared a lie. The lesson: never trust AI outputs without a human filter.

What 'Garbage In, Garbage Out' Means for Your Feed

Atwood nailed it: "Garbage in, garbage out." If you give an AI vague prompts, you'll get generic replies. If your brand voice isn't clearly defined, the AI will default to sterile corporate speak. I once saw a brand's AI-generated Instagram story that sounded nothing like their usual voice—they lost followers because it felt fake.

Here's what you can do:

• Write detailed prompts that include tone examples, brand dos and don'ts, and exactly what you want the AI to avoid.
• Always review AI text for factual accuracy, especially dates, names, and statistics.
• Use AI for first drafts, not final copy—edit until it sounds like a real person.

Ever asked an AI for a caption and gotten something that sounds like a robot wrote it? You can fix that by feeding it 3 of your best past posts as examples.

How to Turn AI into a Reliable Assistant

The jury is still out on whether AI will replace human content creators. But I can tell you: every time a client has gone fully automated, their engagement tanked. On the flip side, clients who use AI as a brainstorming partner—then rewrite everything by hand—see consistent growth.

I could be wrong about this, but my hunch is that social media platforms are starting to penalize generic AI content. Authenticity wins algorithms. If you take away one thing from this, let it be: always test AI outputs before hitting publish. One wrong fact can undo months of trust.